Amazon Cut 7 Times in 2026. Watch 6 Employers, Not 548 Alerts.

Second Talent's tally of the layoffs.fyi database through Sept. 10, 2026 puts Amazon at seven separate 2026 layoff entries, with PayPal and Meta at four each and Samsung, Uber, and Salesforce at three. That is not a news feed. That is a shortlist of six employers you should have been building a private CRM around since February.

Most sourcers still open layoffs.fyi like it is Twitter: refresh, react, blast a template at whoever the tracker just tagged. The math no longer supports that. TrueUp has 548 tech layoff events logged for 2026 at a cadence of about 760 impacted workers a day. You cannot chase 760 people a day. You can, however, run a round-by-round watchlist on six companies that keep cutting on schedule.

Six employers, most of your 2026 poachable pool

A handful of serial cutters generate a disproportionate share of the sourceable 2026 pool, and Amazon alone accounts for a plurality of rounds. Here is the raw concentration from the Second Talent tally of layoffs.fyi through Sept. 10, 2026:

Six companies. Twenty-four rounds between them. Compare that to the 548 discrete layoff events TrueUp logged across the sector this year, and the concentration is the story. US-headquartered companies produced 103,957 of the 128,536 layoffs on layoffs.fyi in 2026, or 81%. Retail leads the tag cloud at 20,200 (mostly Amazon), consumer at 17,544 (led by Meta), hardware at 14,273 (Dell's 11,000 doing the heavy lifting).

The mechanism is boring and it is why alert-driven sourcing has broken down: successive rounds at the same employer create predictable, repeated inventory in known cities. Amazon's Q1 corporate cut showed up in Washington WARN filings as 2,198 positions in the Seattle area, including more than 1,400 in Seattle proper and about 630 in Bellevue. Round 4 will show up in the same two zip codes. Round 5 too. That is a market you can staff a desk against. A one-off alert at a Series B in Denver is not.

760

Amazon's seven rounds are seven different candidate pools

Amazon's 2026 layoffs are not one story. They are seven distinct skill cohorts separated in time, and treating "ex-Amazon" as a single keyword throws away the signal that lets you match a candidate to a client role. Here is what the rounds actually hit:

• January corporate cut. Roughly 16,000 corporate positions worldwide, second major round in three months. WARN filings surfaced 2,198 Seattle-area jobs.
• February, AWS Professional Services. Technical account managers and solutions architects, per field observation from Kore1.
• Early 2026, Alexa AI. Several hundred positions in Sunnyvale and Seattle as Amazon reallocated to its LLM-native Rufus product.
• April, customer service.
• May, third-party seller support (Selling Partner Services).
• July, AGI organization. Unspecified reductions inside the model-customization and post-training group.
• Rolling cuts in Prime Video and Amazon MGM Studios.

An "ex-Amazon SDE in Seattle" from January is not the same candidate as an "ex-Amazon AGI post-training researcher" from July, and pretending they are is why your outreach reply rates keep sliding. The higher-precision filter is division plus month. That is also why plain-English sourcing beats Boolean here: describing what you actually want ("Amazon Alexa AI engineer laid off between January and June 2026, based in Sunnyvale or Seattle, ML systems background") is the exact gap Refolk closes. You describe the person, Refolk returns a ranked shortlist across GitHub, LinkedIn, and the open web without you re-learning search syntax for each round.

The freshest pool goes stale in 60 days

Candidates from a given round are effectively cold by day 60, which flips the poachable-vs-poached math against alert-driven sourcers. Kore1 flagged this directly: AWS Professional Services engineers laid off in February were already 60-plus days active by the time most recruiters engaged. By day 90, the strong ones are placed, and you are sourcing the residual.

In Refolk's index of professional profiles, there are only 114 US-based profiles matching "former Amazon" against SDE and SWE titles. Seattle (5) and Bellevue (3) dominate that cohort, which tracks the WARN filing geography exactly. The second-most-common current employer among that ex-Amazon SDE pool is Google (4), then Microsoft and Oracle at 2 each. That is a thin pool, and it is thin because the cohort that separated in the Q1 rounds has already been picked over.

Now the counterweight. That same index surfaces 1,282 US-based Solutions Architect profiles matching "AWS solutions architect," and AWS itself is still the top current employer (14 of the top-25 sample), followed by CloudFountain, Cloudsfort, VMware/Broadcom, and Pfizer. The derived ratio: for every one ex-Amazon SDE the index surfaces, roughly 11 AWS-badged Solutions Architects are still inside the company. That is the number that should reframe your week.

11:1

Work the survivors, not just the leavers

The highest-leverage move on a serial cutter is warm outreach to the people still inside between rounds. Blind's Amazon forum reliably lights up with internal signaling before every round, which means the survivor cohort knows a cut is coming before layoffs.fyi does. If you talked to them in February when their colleagues went, you have a return conversation in May when the org chart shifts again. The alert-driven sourcer starts cold every round.

This is the round-by-round playbook the six-employer watchlist unlocks:

1. Log the round. Division, month, geography, headcount if disclosed. Amazon Alexa AI, early 2026, Sunnyvale and Seattle, several hundred.
2. Source the leavers inside 30 days. After day 60 you are competing with everyone who set a Google Alert.
3. Source the survivors on the same team. They are the pool for Round N+1, and you already have a reason to reach out.
4. Re-run at each new round on the same employer. Amazon has produced seven so far. Salesforce three, most recently a September cut of close to 1,000 across marketing, product management, data analytics, and its Agentforce AI unit.

AI restructuring is a division-selector, not a company-selector

Challenger, Gray & Christmas reports that AI has been cited in about 23% of all 2026 job cut announcements, the fourth straight month it led stated reasons, with roughly 140,000 tech layoffs in the US so far this year, more than any other industry. May was the worst single month since August 2024, easing in June. But "AI restructuring" as a headline tells you nothing about who to call.

At Amazon it hit Alexa (replaced by Rufus) and the AGI post-training group specifically. At Salesforce it hit Agentforce alongside marketing and data analytics. Sourcers who filter serial cutters by which sub-team got cut this round get a clean skill match that "ex-Amazon" as a keyword never will. An Alexa NLU engineer displaced by Rufus is a plausible hire for a voice interface team; a Selling Partner Services PM is not, and both show up under the same LinkedIn filter.

The underexploited cutters: PayPal and Salesforce

Everyone piles onto Amazon alerts, which is exactly why the ex-Amazon SDE pool is thin. PayPal has four 2026 rounds and Salesforce three, and neither draws the same recruiter crowd. Refolk's index returns only four US profiles for "former PayPal engineer" against senior SWE titles, a thin and largely uncontested pool where a serial-round watch pays off disproportionately. Salesforce's September round hit Agentforce AI, marketing, product management, and data analytics, which is a very specific skill mix and a good test of whether your sourcing stack can filter on team plus month rather than company plus title.

The broader cutter list beyond the top six is worth a watch too: Atlassian has cut roughly 1,600 roles (about 10%), Block roughly 4,000 (about half its workforce), Etsy about 220 (12%, mostly product and engineering), Dell 11,000, and Intel 27,058 cumulatively across 2025 and 2026. None of these are Amazon-scale, but Block and Atlassian in particular concentrate a lot of senior engineering into a small number of geographies.

FAQ

Should I stop using layoffs.fyi? No. Use it as a backfill against your watchlist, not as your primary feed. The daily digest is noise at 760 impacted people per day and 548 events year to date. The value in the underlying database is that it lets you count rounds per employer, which is how you built the six-company shortlist in the first place.

How do I identify the survivors on a team that just got cut? Start with the division and geography from the WARN filing or press coverage, then pull current employees at that employer with the matching team keyword and location. For the Amazon January round that means Seattle and Bellevue corporate; for Alexa it means Sunnyvale and Seattle with voice, NLU, or speech in the title. Describing the survivor cohort in plain English to Refolk is faster than assembling a nested Boolean, especially when you are re-running the same query month after month.

Is 60 days really the cutoff for a fresh pool? It is a working rule from Kore1's field observation on AWS Professional Services engineers in the February round, not a law. The stronger the engineer and the tighter the market for their skill, the faster the clock. For niche cohorts like AGI post-training researchers, assume closer to 30 days. For generalist corporate roles from the January 16,000, you have longer, but competition is proportionally worse.

Why only six companies on the watchlist? Because 24 rounds across six employers is already more work than most desks actually do rigorously. Add a seventh or eighth from the adjacent list (Atlassian, Block, Dell, Intel) only when you have a client brief that maps to one of their divisions. Watchlist discipline beats coverage.